Growing the Vote: Majority Party Whipping in the U.S. House, 1955-2002 C. Lawrence Evans Department of Government College of William and Mary Presented at the 10 th Annual Congress and History Conference, Brown University, June 9-10, 2011
Growing the Vote:
Majority Party Whipping in the U.S. House, 1955-2002
C. Lawrence Evans
Department of Government
College of William and Mary
Presented at the 10th
Annual Congress and History Conference,
Brown University, June 9-10, 2011
!
"!
Growing the Vote:
Majority Party Whipping in the U.S. House, 1955-20021
This paper reports on a broader study about partisan coalition building in the U.S.
House and Senate, with an emphasis on the role of the whips and other party leaders prior
to votes on some of the most significant legislation considered on Capitol Hill since
World War II. By “whip” and “whipping,” I am referring both to positions of formal
leadership within the Congress (each party within each chamber designates members as
whips) and also to the myriad of tactics that party leaders employ to advance the party
agenda on the floor (gauging the emergent preferences of members, lobbying potential
defectors from the party program, altering bills or providing special favors to build
support, and so on). Here, my focus is on majority party whipping in the House from
1955 to 2002, by all accounts an era of remarkable institutional transformation within the
chamber.
In particular, I consider five main questions. How active were the majority party
whips during this period and what factors account for any changes that may have
occurred over time? What kinds of questions and issues have House majorities typically
singled out for disproportionate whipping? Generally how large is the initial base of
support that party leaders can count on at the beginning of floor fights on major bills?
Have certain whips been especially effective at mobilizing member support for the party
program and why? Does the partisan majority generally carry the day on whipped
matters, and if so, are these victories accomplished without major substantive
concessions on the floor? How might the answers to these questions inform ongoing
debates among scholars and others about the role and impact of the political parties in
Congress?
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
"!Portions of the evidence gathered for this paper (materials for the 1955-86 period) benefited from the
support of the National Science Foundation (Award SES-0417759), The Carl Albert Center of the
University of Oklahoma, and the Thomas Foley Institute of Washington State University. More recent
evidence (primarily materials for the 1990s) was secured in part because of support from the Roy R.
Charles Center and the Thomas Jefferson Program in Public Policy of the College of William and Mary, as
well as the Newton Family Trust. My most significant intellectual debt is to W. Lee Rawls. The title
phrase, “growing the vote,” is from DeLay (2007).
!
#!
The analysis of this paper is divided into six sections. First, I provide background
about the whip function and the nature of my evidence. Section 2 is an exploration of the
magnitude and composition of the whipped agenda for the House majority party over
almost fifty years of recent House history. Next, I consider evidence from whip counts
(private polls conducted by party leaders prior to important votes) to gauge the magnitude
of the coalition-building challenges that have confronted House majorities on the floor,
and also their relative success in mobilizing support, or “growing the vote.” Section 4 is
an analysis of the outcomes of whipped questions for the majority party. Section 5
explores the whip count behavior of centrist members of the House, which is especially
useful for evaluating theories of lawmaking. Concluding comments about the
implications of this research are offered in Section 6.
1. The Whips as Evidence
In the House, the office of party whip is the third ranking position for the majority
party (following the Speaker and majority leader) and second ranking for the minority
(after the minority leader). The position of Republican whip was created in 1899 by then
Speaker Thomas Brackett Reed, and two years later the minority Democrats established
an analogous office on their side of the aisle. According to reliable media accounts and
the contemporaneous observations of members, the early whips served as key lieutenants
to party leaders, conducting tallies of the preferences of fellow partisans before major roll
calls, ensuring that supportive members were present on the floor for important decisions,
and making arrangement for paired votes for absent members (Clark, 1920).
As the House parties became more internally divided after the 1910 revolt against
Speaker Joseph Cannon, legislative authority shifted from party leaders to the committee
system and the role of the whips became more circumscribed for a time. During the New
Deal era, however, the House Democratic Whip, Patrick Boland of Pennsylvania, was a
valued and highly effective ally of the Roosevelt administration. In the early 1930s, both
House parties also expanded their whip operations to include networks of assistant,
regional, and zone whips with geographic responsibilities. In 1933, for example, the
House Democratic whip operation included 16 members, with one zone whip responsible
for working with Democratic members from New England, another fulfilling this role for
!
$!
the large Texas delegation, and so on (Evans and Grandy, 2009). The size of the House
whip systems did not change all that much until the late 1970s, when the Democrats and
then the GOP steadily expanded the number of members with some role in the party whip
operation (Sinclair, 1995, Dodd 1978). By the mid-1990s, in a House increasingly
polarized along party lines, fully 100 Democrats were whips of one kind or another,
while on the Republican side of the aisle about 70 members had formal roles in their
party’s whip operation.
Although the size and organizational complexity of the two House whip networks
have changed dramatically over time, the basic ingredients of effective whipping have
been in place for decades. Among other duties, the whips work with other members of
the leadership to formulate and implement party strategies; canvass the views of members
of the relevant party rank and file prior to major floor actions; lobby individual members
to stay loyal to the party program; manage and run regular meetings of the extended party
whip organizations; provide members with policy and political information about major
bills likely to appear on the floor agenda; communicate regularly with member offices
about the upcoming floor schedule, especially the timing of votes; and act as public
spokespersons for the party.
For these reasons, the whip networks of the House provide a unique vantage point
for evaluating the tactics and impact of the congressional parties. There is considerable
dissension among scholars and others about the causes and consequences of the rampant
party polarization that characterizes the contemporary Congress and concerns so many
citizens (Sinclair 2006, Theriault 2008). Is congressional partisanship primarily a
product of the policy preferences of individual members, which in turn are shaped by
constituency and other pressures rooted in the electoral environment (Krehbiel, 1998)?
Or are party leaders and other internal characteristics of the Congress responsible for
much of the rampant partisanship that has emerged on Capitol Hill since the 1970s? To
the extent that it exists, is party power mostly procedural and centered at the agenda
setting stage (e.g., Cox and McCubbins, 2005), or can leaders systematically whip their
fellow partisans into line on the floor (Rohde, 1991, Aldrich and Rohde, 1998)?
Roll call data, of course, are the most accessible and commonly used form of
evidence for analyzing congressional decision making, but the limitations of the roll call
!
%!
record for identifying the causes and consequences of partisan behavior are well-known
and fairly obvious (Kingdon, 1973, Jackson and Kingdon, 1992, Smith 2007). For one, a
straight party-line vote on the House floor is not necessarily an indicator of leadership
pressure. Plausible alternative explanations include the ideologically divergent
constituencies represented by Democratic and Republican members, the lobbying
effectiveness of interest groups that share agendas with one or the other party, pressure
from the White House on members of the president’s party, and even the different
political philosophies held by Democratic and Republican members. Without systematic
information about the positions and preferences of members prior to the voting decision –
and about the lobbying tactics employed by party leaders and other coalition builders –
there is simply no way to know for sure. An in-depth exploration of party whipping can
provide just that evidence.
The whips have received only limited attention from political scientists,
historians, and even journalists, however, because most of their activities are hidden from
public view (but see Ripley 1964, Dodd 1978, and Burden and Frisby 2004). The
canvassing of member viewpoints before major floor votes, for instance, can only be
effective if members know that their personal communications with the leadership will be
kept private and not shared with their political opponents or voters. Confronted by an
upcoming floor fight on a significant issue, party leaders also will not want their strategic
plans shared with leaders from across the partisan aisle. Indeed, in fall 2009, there was
an uproar within the House Democratic Caucus after party liberals shared with journalists
the results of an internal whip count regarding the so-called public option then under
consideration for inclusion in the party’s health care reform bill.2
Fortunately, the archived papers of former congressional party leaders include
extensive records of party whip operations from the mid-1950s to the early 2000’s, and a
nearly comprehensive paper trail is available for the majority party. Not surprisingly,
there are some gaps. Evidence about majority party whipping by House Democrats is
unattainable for most of 1976 and 1991, and for the entirely of 1987, 1988, and 1990.
For the Republican majorities of 1995-2002, I was able to comprehensively identify the
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!2 Mike Soraghan, “Whip count shows Democrats lack votes on ‘robust’ public option for healthcare,” The
Hill, October 22, 209.
!
&!
issues on which the majority whips were active, and detailed evidence about the emergent
positions of individual members has been gathered for 1997-2002, but only sporadic
information at the member level is available for 1995-96.3 Still, approximately 30,000
pages of records have been reproduced and examined as part of my broader research
about the whips, touching on much of the major legislation considered on Capitol Hill
during the second half of the twentieth century.
2. Majority Whip Activity
One useful indicator of party coalition building on the floor is the number and
substantive focus of party whip counts. As mentioned, whip counts are essentially
private polls conducted by House and Senate party leaders prior to major floor votes.
Although there is some variance across parties, chambers, and over time, leaders
generally use these polls to categorize their fellow partisans as “yes,” “leaning yes,”
“undecided,” “leaning no,” “no,” or nonresponsive in some way on the upcoming matter.
In the House, the majority whip process usually begins with a request from the majority
leader, usually in consultation with the Speaker and other top party leaders, for a count of
the leanings of rank-and-file members of the party on a matter pending before the
chamber. The whip communicates the request to members of the extended whip system
for the party (called zone whips by the Democrats and assistant or deputy whips by the
GOP). These individuals in turn contact the members from their party over which they
have been assigned responsibility and ask for responses to one or more specific questions
that relate to the targeted matter. For the Democrats, whip zones have always been based
on geography. The nature of the division of labor within the House Republican network
has changed over time, and in the late 1990s was based on both geography and the
personal ties of particular members. Especially for the Democrats, the questions being
polled tend to be phrased so that a “yes” response is in favor of the position of the
leadership.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!3 For various reasons, information about whip behavior for the House minority party is only available for
the 1970s and 1989-1993. Archival evidence about party whipping in the Senate is mostly confined to the
late 19980s and 1995-2002. Evans and Grandy (2009) and Evans (2011) provide full information about the
archival sources used for this project.
!
'!
As they hear from their colleagues (directly or through staff), the zone or assistant
whips report back to the party whip with their tallies, singling out particular members as
potential problems or in need of contact or persuasion from the leadership. Typically,
members of the whip organization meet with the relevant party leadership and (perhaps)
leaders from the committee or subcommittee of jurisdiction to plot strategy and
coordinate the pre-vote lobbying effort. These follow-up efforts often are referred to as
“clean-up” and can be pivotal to the outcome.
Figure 1 summarizes the activity levels of the House majority whip operations,
1955-2002. The figure traces the number of polled questions that the majority leadership
put to its members each year. If the leadership requested that the zone or assistant whips
ask their members three separate questions dealing with a single bill (perhaps about the
rule, an expected floor amendment, and the vote on final passage), then the request is
treated as three polled questions. The missing entries are for years for which
comprehensive whip count data for the majority party are unavailable. Overall, the
evidence indicates several instructive trends.
Obviously, the level of majority whip activity has risen substantially over time.
For the Democrats, the average number of whip counts per year for 1955-72 was about
9.3. During 1973-1982, the average rose to 29. During 1983-94, the average increased to
a little over 35. For the Republican majorities of 1995-2002, the average number of whip
counts per year was 55.5.
Although the number of whip polls has increased over the years, there were sharp
spikes for the Democrats in the late 1970s and for the Republicans in the late 1990s. The
well-known increase in partisan polarization from the 1970s to the late 1990s was
probably responsible for some of the long-term growth in whip activity, especially the
GOP spike during 1995-2000. Moreover, compared to prior Democratic majorities, the
Republican majorities of 1995-2002 were relatively small. The modest size of the GOP
margins during these years, combined with the high levels of partisan polarization, meant
that the party needed to hold onto the votes of almost all of its members to win on the
floor. One consequence was a highly active whip operation, which is reflected in the
hundreds of counts that GOP Whip Tom DeLay supervised during the period.
!
(!
The surge in whip activity among the majority Democrats during the late 1970s,
in contrast, occurred during a time of relatively low partisan polarization and sizable
majority party margins (almost 300 members). Why was the Democratic operation so
active during those years? Prior to the Legislative Reorganization Act of 1970, recorded
votes were not conducted in the Committee of the Whole, which is the parliamentary
device through which the chamber considers most amendments to legislation. With the
inception of recorded voting in the 1970s, the number of floor amendments grew
markedly, in part because the minority GOP sought to force the majority Democrats to
cast public votes on politically divisive issues that could hurt them at home. Confronted
with an avalanche of amendments and votes and a caucus that was large but divided, the
Democratic leadership significantly stepped up its whip activity.4 By the early 1980s,
Democratic members had convinced Speaker O’Neill to clamp down on the floor
amendment process via the use of more restrictive rules. For a time, the number of
majority whip counts fell, before again rising because of the burgeoning partisan
polarization within Congress.
Overall, about 70 percent of the over-time variation in majority whip activity can
be explained with just two factors – the degree of partisan polarization within the House
(as measured by the distance in mean DW-NOMIINATE scores for Democrats and
Republicans) and the size of the chamber workload (as reflected in the number of
recorded votes per Congress).5
Table 1 provides summary information about the kinds of questions that were
subject to majority whip activity, with 1955-2002 divided into four discrete periods that
existing scholarship indicates are meaningful for understanding party politics. The first
period, 1955-1972, covers the last decade or so of the “Textbook Congress,” in which the
Democratic majority was often deeply divided along sectional lines, especially on matters
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4 Moreover, the election of Jimmy Carter as president in 1976 meant that Democratic leaders were
responsible for shepherding the administration’s legislative agenda through Congress, reinforcing the need
for accurate intelligence about the evolving positions of members and for effective lobbying on the floor. 5 The number of majority whip counts was regressed (using both ordinary least squares and negative
binomial regression) against polarization, the number of roll calls, the minority party’s share of the
membership, an interaction term that is the product of polarization and the minority proportion, and an
indicator variable taking the value of one if the House majority party also controlled the White House.
Only the polarization measure (when included separately and without the interaction term) and the number
of roll calls produced statistically significant parameter estimates, and the impact of the number of votes
was only slightly larger than the impact of polarization.
!
)!
of civil rights and race, and committees had significant autonomy vis-à-vis the majority
leadership. The next period, 1973-82, begins with the inception of recorded voting in the
Committee of the Whole and the resulting surge in floor activity. Here, as a result of the
1960s enfranchisement of black Americans and the shift among Southern conservatives
toward the GOP, the longstanding sectional cleavage within the majority Democratic
Caucus grew substantially less important. But the large Democratic majorities of the
1970s were still deeply divided on the major issues of the day, the chamber was
characterized by rampant member individualism, and at least during 1981-82, the party
confronted an effective Republican president with an ambitious legislative program
(Sinclair 2006). Following the Democratic Party’s strong performance in the 1982
midterm elections, however, the House majority leadership stepped up its opposition to
the Reagan policy program and party unity within the Democratic Caucus began to rise
(Rohde 1991). This turn toward partisan polarization in Congress continued through
1994. The 1995-2002 period, of course, marked the emergence of a Republican majority
in the House for the first time in forty years. According to some scholars, the GOP
majorities of the mid to late 1990s are the closest the House has come to party
government since the turn of the 20th
Century (Aldrich and Rohde 1998).
In Table 1, the top row includes polled questions about the passage of entire bills
or resolutions and the broad issue stances of members. The next category for conference
matters includes questions about the adoption of conference reports and related motions,
such as motions to instruct conferees during bicameral bargaining. The third row in the
table is for questions about amendments, portions of legislation, and other discrete policy
alternatives. Since the motion to recommit legislation to committee, with or without
amendatory instructions, is a valued minority party prerogative within the House, such
motions are treated as a separate question category in Table 1. The category for
procedure and strategy is mostly compromised of questions about special rules and
previous question motions on special rules, but also includes a range of other tactical
questions that are primarily about process, rather than the substance of legislation. The
final category is for questions about attempts to override a presidential veto.
As Table 1 indicates, overall just over half of the polled questions put to members
by the majority leadership, 1955-2002, dealt with entire measures or general policy
!
*!
matters. About 20 percent pertained to amendments or other alternatives, about 13
percent concerned procedure and strategy, and slightly less than seven percent were about
conference matters. Relatively few whip counts dealt with motions to recommit or veto
overrides.
Still, as expected, there is noteworthy variation across time periods. During 1955-
72, prior to the inception of recorded votes on amendments and the dramatic rise in floor
amending activity that ensued, the percentage of majority whip activity focusing on entire
bills and broad policy matters was very high, over 70 percent, and the proportion of whip
activity targeting the amendment process was relatively low. Not surprisingly, when the
majority leadership began to rely on restrictive amendment procedures on the floor
during the late 1970s, the proportion of whip activity allocated to procedural matters
basically tripled. Indeed, during 1983-94, about one in five whip counts concerned
procedure.
Whip attention to veto overrides was most prevalent during the 1970s, especially
in 1973-76 when Republicans controlled the White House and Presidents Richard Nixon
and (especially) Gerald Ford made regular use of the veto pen to block initiatives
champion by the majority Democrats. Indeed, if we focus exclusively on that four-year
span, 13 of the 88 polled questions (about 15 percent) dealt with veto overrides.
Interestingly, during 1997-2000, when the Republican House majority regularly squared
off against the Democratic administration of Bill Clinton, not a single one of the several
hundred whip counts conducted by GOP Whip Tom DeLay dealt with a veto override
attempt. The reasons are straightforward. The Democratic majorities of the 1970s were
relatively large and cross-partisan coalitions were fairly common, making the successful
override of a presidential veto potentially attainable. In the late 1990s, in contrast, GOP
leaders had small margins and on veto overrides they could count on highly unified
opposition from congressional Democrats. DeLay and other Republican leaders were
reluctant to waste their time and political capital whipping for override attempts that
almost certainly would fail.
Additional insight about the “whipped” agenda can be derived from comparisons
with the contents of the roll call record. Table 2 juxtaposes the kinds of questions
targeted for majority whipping with the contents of the questions subject to a recorded
!
"+!
vote on the House floor during the relevant years. Of the more than 1,000 questions
polled by the House majority party, 1955-2002, for which there are archival traces,
almost 80 percent can be linked directly to a roll call that occurred on the House floor,
and the lion’s share of these votes were nonunanimous (that is, neither side received
fewer than 10 percent).6 For the population of House roll call votes that occurred during
the period for which majority whip poll data are available, I identified which votes linked
to whip counts and then compared the composition of the two sets of questions –
nonunanimous votes with polls and nonunanimous votes that were not whipped.7 The
cell entries in Table 2 are the percentage of the relevant agenda (whipped votes versus the
others) that fell in the aforementioned question categories. For the 1955-72 period, for
example, roughly 61 percent of polled votes dealt with House passage, while only 35.52
percent of the other roll calls were in this category.
Overall, the kinds of questions that were whipped were fairly representative of the
broader agenda, albeit with certain exceptions. The disproportionate emphasis on House
passage is also apparent in Table 2. While this category accounts for about 50 percent of
the polled votes, it comprises just 21.3 percent of the other nonunanimous roll calls. At
least since the early 1970s, questions related to the floor amendment process are less
prevalent among polled votes relative to the others.
In Table 2, to facilitate comparisons between whip counts and the roll call record,
the category for procedure only includes special rules and previous question motions on
special rules.8 Under the Democratic majorities of the 1970s and 1980s, as the majority
leadership made increased use of restrictive procedures to limit minority party
participation on the floor, these motions were much more prevalent on the polled agenda
than they were in the roll call record writ large. During the pre-1970s “Textbook” era, in
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!6 At this point in the analysis, whip counts are linked to votes even if there were certain modifications to the
underling text between the time of the poll and the occurrence of the vote. For example, if the majority
leadership tallied member positions about the bill as reported by committee, but a major amendment was
adopted on the floor, the count is still linked to the final passage vote because our interest for now is in the
kinds of matters that evoke whip attention, rather than the relative success of the leadership’s lobbying
effort. 7 In assigning votes to question categories, I relied on Rohde, David, “Political Institutions and Public
Choice House Roll-Call Database,” Duke University, Durham, NC, 2010. 8 With the exception of special rules and previous question motions, it can be difficult to draw systematic
linkages between the sorts of questions tagged as “procedural” or “strategic” in Table 1 and particular vote
categories in Rohde 2010.
!
""!
contrast, amendment rules tended to be relatively open and the associated votes often
were not partisan, reducing the call for partisan whipping. And by the mid to late 1990s,
the expectations for party loyalty on rule votes may have been so strong that whipping
typically was not needed for the majority party to prevail.
Table 3 repeats the analysis of Table 2, but now votes (both polled items and
other roll calls) are categorized by policy area, rather than question type.9 Once again,
there are important similarities. While items pertaining to the economy, budget, and
taxation comprise about 20 percent of polled votes, they are only responsible for about 10
percent of the other nonunanimous roll calls. The differences between the two agendas
largely derive from tax items and budgetary matters, which were significantly more
common among polled items than they were among the other votes. Interestingly, the
disproportionate emphasis on tax policy in the whipped agenda characterizes the
Democratic majorities from 1973 to 1994, but not the Republican majorities of the 1990s.
In part, the reason may be that tax votes had become such a regular feature of the roll call
agenda by the early 1990s. But another factor is probably the substantial cohesion that
had emerged among Republicans on tax issues, which reduced the need for them to
engage their whip operation.
Table 3 also reveals noteworthy changes over time for appropriations bills, the
annual spending measures that can create important managerial challenges for whichever
party is organizing the chamber. Over 1955-2002, the proportion of roll call votes that
occurred on appropriations bills grew substantially, from under 20 percent to almost one-
third. Prior to the 1980s, party whip counts were relatively uncommon on appropriations,
reflecting the bipartisan norms that structured the spending process during those years.
Beginning in the 1980s, but especially during the GOP majorities of the 1990s,
appropriations issues became a central feature of the whipped agenda. Indeed, during
1995-2002, over 35 percent of majority party whip counts in the House targeted
appropriations bills in some fashion.
Interesting, this increase in polling activity did not target defense issues or the
Labor-HHS spending bill, which touch on hot-button issues that often divide the two
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!9 Here, I relied heavily on the categorization scheme developed by Rohde and his colleagues. For details,
consult the codebook for Rohde (2010).
!
"#!
political parties. Instead, the major disparities between the whipped agenda and the
larger roll call record appear to center on appropriations for foreign operations, the
legislative branch, the District of Columbia, and the Treasury-Postal bill. Regardless of
which party has majority control of the House, it can be difficult to build floor majorities
for these measures because they lack strong domestic constituencies and often are
inviting targets for position taking by individual members. Spending on foreign aid, for
instance, is seldom popular with U.S. voters. The majority party often whips the previous
question motion on the rule for the Treasury-Postal bill because defeat of that motion is
the primary mechanism through which self-styled populists can cut the salaries of
members of Congress. Similarly, the appropriations bill that funds the legislative branch
is a highly inviting amendment target for members seeking to position themselves as self-
sacrificing reformers. The failure to move these measures through the House in a timely
fashion, however, can create bottlenecks in the floor schedule, endangering other aspects
of the majority party program. Moreover, gridlock in the annual appropriations process
can result in a partial shutdown of visible federal programs, creating political problems
for majority party members at home. Much of the disproportionate whip activity on
appropriations matters, in other words, derives from the managerial responsibilities
associated with majority status, irrespective of which party is organizing the chamber.
3. Whip Polls and Votes
The remarkable scope of majority whip activity is suggestive of majority party
influence in the legislative process. On the other hand, high levels of whip activity are
also a possible signal of significant weaknesses in the majority coalition. Within both
leaderships, there are significant opportunity costs to engaging the whip system. Political
capital is not unlimited and allocating significant effort to one item reduces the stock of
resources, goodwill, and obligation that leaders can draw on for other aspects of the
policy agenda. Moreover, the willingness of rank-and-file members to make concessions
to their leadership, that is, to downplay their personal priorities and instead promote the
collective interests of their party, is likewise limited. For these reasons, the majority
leadership should devote its attention and political capital to items and initiatives that (1)
have at least the potential to unify the party rank and file; (2) where the outcome is still in
!
"$!
play; and (3) the underlying issue is a significant element of the party program or
message agenda. As mentioned, since the public holds the majority party
disproportionately responsible for managing day-to-day business within the chamber, the
majority leadership also will need to build winning coalition on “must-pass” elements of
the legislative agenda like appropriations bills.
As a result, in gauging the impact of partisan institutions like the whip networks,
we need to consider the nature of the challenge confronting the leadership and its relative
success or failure in meeting that challenge. One way to gauge this challenge on the floor
is to examine the initial base of support for the party position at the beginning of the whip
process.10
Consider Figure 2, which compares the percentage of majority party members
supporting the party position on whip counts with party cohesion on roll calls
(nonunanimous and party-line). Unfortunately, evidence about the poll responses of
individual members is not available for all of the whip counts for which there are
meaningful archival traces (the text of the polled question, the date, and other general
information about the lobbying process may be available, but not detailed information
about the emergent positions of individual members). Moreover, it was not unusual for
there to be so many nonresponses or missing positions that the count appeared to be
incomplete, or the leadership only polled a small number of members who were viewed
as “in play” on the matter. Including these items might distort aggregate measures of
party support across whip counts and they are dropped from this portion of the analysis.11
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!10
One common question about whip counts is whether the responses of members can be trusted. Are there
any incentives for them to overstate their opposition to the party program, potentially securing favors in
exchange for their support? Four points are worth mentioning in response. First, the whip process is a
“repeated game” and members develop reputations. There are incentives for them to be truthful. Second,
congressional leaders generally know a lot about the constituencies of rank-and-file members and can be
very difficult to fool. Third, in a sense it does not matter. If a member claims that she will oppose a bill or
amendment unless she receives some concession, then that essentially becomes her position and the polled
question and the concession are for all practical purposes inseparable. Fourth, and most important,
participants in the whip process believe that whip poll responses are accurate, which is precisely why they
base strategic decisions on the results. 11
A count is treated as incomplete or partial if response data of some form is missing for more than 60
members. During the Republican era, Tom DeLay made regular use of targeted whip checks and these
items were denoted as such in office records.
!
"%!
Here, party support on a whip count is the sum of “yes” and “leaning yes”
responses divided by the total number of responses.12
As the figure demonstrates, initial
party support on whip counts and party cohesion on votes tend to move together and have
increased markedly over time. The only significant exception is the 94th
Congress, 1975-
76, and this may be largely an artifact of the record keeping practices of the leadership at
the time.13
In the figure, party support on whip counts is systematically lower than roll call
cohesion in part because the number of members switching from opposition on the count
to support on the vote usually exceeds movement in the opposite direction. But the
primary reason is that so many lawmakers respond as “undecided” on whip counts or
otherwise decline to make a firm commitment one way or the other on the polled matter
(they may respond as “ no comment” or provide a substantive response that the whips
cannot record using one of the standard response categories). As was the case for
partisan behavior on roll calls, the size of the base on majority party whip counts
(calculated as the sum of members responding as yes or leaning yes) has increased
steadily and significantly over time. Much of the heightened party polarization in the
House, in other words, appears to have been apparent near the beginning of the process of
endgame lobbying on the floor.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!12
Nonresponses and members who were not included in the poll are excluded, as are the occasional
response that is illegible. Typically, the archival records of whip operations includes multiple drafts of a
count, with the positions of some members changing from draft to draft as the lobbying process continues
and the roll call nears. Indeed, for some issues five or more lists of the results may be compiled in the day
or two before the relevant vote. The main changes from earlier to later drafts typically are a reduction in
the number of nonresponders. To simplify matters, when the positions of members change across drafts, I
coded the position that maximizes the level of response ambiguity. If a member is “undecided” on one
draft and “leaning yes” on the next, the position is coded as “undecided.” If there is a tie in terms of
distance from undecided (e.g., shifts form “no” to “yes,” or from “leaning no” to “leaning yes”), then I
coded the response that was furthest from the leadership position. A nonresponse is coded if and only if the
member is nonresponsive across all drafts. 13
In contrast to the rest of the whip data, evidence for the 94th
Congress could not be obtained from the
personal papers of the relevant whip, John McFall, D-CA. For some reason, there are no traces of
McFall’s leadership activities in his personal records. Instead, evidence about whip activity during 1975
and the first two months of 1996 was obtained from the papers of Carl Albert, who was speaker at the time
and who kept records of the results of party whip counts. In contrast to the other whip records gathered for
this project, these materials may predominantly reflect member positions at the very end of the whip
process, and immediately prior to the relevant vote. This may explain why the level of party support on
these counts was so similar to that apparent in the roll call record. Larry Dodd, who was an APSA
Congressional Fellow in the whip’s office during the 93rd
Congress, gathered whip data for McFall’s first
Congress in the position and kindly shared the information with me.
!
"&!
Table 4 provides more detailed information about the responses of members on
majority whip polls, this time broken down by the four periods. During the waning years
of the Textbook Era, 1955-72, slightly more than 56 percent of members pledged to
support the party program on whip counts and another 2 percent were leaning that way.
About 13 percent were undecided, 11.57 answered “no,” and another one percent was
leaning that way. The category for “other” (various forms of nonresponse) included
15.51 percent of the observations. Looking across periods, the proportion of undecideds
increased somewhat during the 1970s and 1980s, before dropping back to 11.25 percent
during the years of the GOP majority. The percentage answering no or leaning no
declined from about 12 percent during the first two periods to about eight percent during
1983-94 and 1997-2002. Consistent with Figure 2, the party base on whip counts has
increased markedly over time.
As mentioned, the “other’ category includes a range of nonresponses. The
proportion of observations in this category declined from 15.51 percent and 11.95 percent
during the 1955-72 and 1973-82 periods, respectively, to about seven percent in the
1980s and 1990s. Two ingredients of the “other” category merit additional attention,
zone aggregates and “if needed” responses.
Especially during the 1960s, certain geographic zones, mostly from the south,
refused to provide the Democratic leadership with information about the positions of
individual members and instead reported just the totals (e.g., “three are yes,” “four are
undecided,” “11 no,” and so on). Zones that included the Carolinas and especially the
Texas delegation mostly employed this tactic. The recollections of participants are not
completely consistent about the underlying motivation. One view is that Texas
representatives in particular felt like they had strong personal ties to the Johnson White
House and preferred to deal directly with the administration. Another view is that Texas
representatives and other southerners knew that the Democratic whips would share their
information with the White House and these members did not want to be subjected to the
famed “Johnson treatment” of intense personal lobbying.14
There probably is some truth
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!14
For example, Gary Hymel, the top aide to Hale Boggs as whip and also a senior staffer to Thomas P.
“Tip” O’Neill during his speakership expressed the first view to me. The second perspective was
articulated in a personal interview with former Speaker James Wright, who was a member of the Texas
delegation during much of the Textbook Era.
!
"'!
to both explanations. But use of the tactic by Democratic zone representatives primarily
occurred in the 1960s and early 1970s, and would not have been tolerated in the more
polarized houses of the 1980s and 1990s.
On occasion, members tell the whips that they would rather vote against the party
position, but are willing to stay loyal on the roll call “if needed.” Usually, such
commitments are made late in the whip process and are not formally recorded in whip
checks. Instead, a member will respond as no or leaning no to a succession of inquiries,
as the roll call nears the leadership will inform the member that his or her vote may be
pivotal, and at the last minute the member may switch from opposition to support.
However, members occasionally adopt the “if needed” position early enough in the whip
process for the commitment to be formally recorded. Although such occurrences are
relatively rare, their incidence over time is informative about the importance of partisan
ties. If we stay focused on completed whip counts, then there were 94 clear-cut instances
in which a member was listed as “if needed” on majority whip counts, and more than
two-thirds of those instances were from the Republican majorities of 1997-2002. Indeed,
on the GOP whip counts, five or more members told the leadership that they would be
with the party “if needed” on 23 polled questions, and in several instances more than a
dozen members responded in this fashion.
The Base and Majority Pickup
Percentages and support rates are informative, but to properly gauge the coalition
building challenges that confronted majority party leaders and their relative success in
meeting them, what really matters is the number of votes. A 70 percent support rate for
the 300-member partisan majorities of the late 1970s, for example, would translate into
210 votes, which is close to a chamber majority, while an analogous support rate for the
narrow Republican majorities of the 1990s would translate into just 160 votes and a
substantial vote-growing challenge.
Since our interest now is in directly comparing the size of the base and vote
gathering success on discrete items, let’s focus again on the polled items that could be
directly linked to votes. And once again, let’s distinguish between votes that are party
line and all nonunanimous roll calls. For each period, the average size of the majority
!
"(!
party base (yes plus leaning yes responses) on the polled items is provided in the first two
rows of Table 5 (one row for counts that link to nonunanimous votes, and the other for
counts that link to party-line votes). Notice that the average base does get larger over
time, but its absolute size is smaller during the Republican majority, reflecting their
narrow margins. And not surprisingly, the base tends to be slightly larger on the party-
line items.
In the table, the next two rows are for the majority party pickup, which is the
difference between the number of majority party members voting with the party on the
relevant roll call and the size of the base. Interestingly, the average size of the pickup
declined over time for the Democratic majorities, before rising somewhat under the
Republicans in the last period. Over the decades of Democratic rule, in other words, the
size of the base grew, but average pickup fell, illustrating how increased polarization
within the House made floor decision making more predictable and less fluid. The
smaller base and slightly larger pickup levels for the Republican majority (in comparison
with the Democratic majorities of 1983-94) appear to reflect both the small GOP margins
and effective vote gathering by Tom DeLay on the floor.
Expectations about the likely level of support on the underlying matter from
members of the minority party are also a factor in the majority whip process. Party whips
seldom approach members of the opposite party to ask about their views, but allied
interest groups conduct their own canvassing and regularly share political intelligence
with the whips. Alternatively, the whips may approach a member of the other party
indirectly, though a member (of the whip’s party) who has personal ties to the targeted
individual. Indeed, lists of likely supporters from across the aisle are common in the
archival record, and for some measures, such as trade votes, systematic counts may be
conducted by both parties, the White House, and five or ten outside groups covering the
entire House membership. One reasonable approach to estimate the prospects for cross-
partisan support on a polled item is to simply look at the average number of minority
party members who ended up voting with the majority party on the relevant roll call, and
this information is provided in the bottom two rows of Table 5. Not surprisingly, for the
questions that produced party-line votes, the average level of minority party support has
steadily declined over time, reaching a low of 21.64 members during the Republican
!
")!
period. For all nonunanimous votes, in contrast, the average level of minority support
has been fairly stable over time, even increasing somewhat in 1997-2002.
If for party-line items we sum the appropriate rows, then base plus minority party
support was roughly 181 members in 1955-72, 192 in 1973-82, 208 in 1983-94, and 194
in 1997-2002. The gap that needed to be filled to achieve the generic 218 votes was thus
about 37 members for the Textbook period, 26 for the 1970s transition, just 10 members
for the Democratic majorities of 1983-94, and 24 for the Republican years. Interestingly,
the mean pickup closed this gap for every period but the first, where the gap and pickup
were essentially the same. To properly compare the accomplishments of the whips,
however, party size still needs to be factored in.
The results of a more systematic analysis of partisan vote gathering that includes
majority party size are reported in Table 6. For all completed whip counts on which the
majority took a clear position and for which there exists a nonunanimous roll call, I
regressed the number of majority party votes for the party position against the size of the
base (yes plus leaning yes responses on whip counts), the size of the majority party, and
indicator variables for the periods, with 1955-1972 as the omitted category. The
regression was repeated without the period indicators and instead with analogous
indicator variables for the individuals who served as majority party whips in the House,
1955-2002. (Dates for their tenures as majority whip are denoted for reference.) Here,
Carl Albert, Democratic whip from 1955 to 1961, serves as the omitted category. In the
second equation, Tony Coelho, D-CA, and William Gray, D-PA, are necessarily captured
by a single variable because only one year of comprehensive whip count data is available
for their combined tenure in the office: The year is 1989, first session of the 101st
Congress. During that session, Coelho was whip for the first six months and Gray for the
last six months, and for practical reasons it is not possible to separate out their tenures for
analysis.
In both regressions, the parameter estimates for the initial base and party size have
the expected signs and are statistically significant.15
Not surprisingly, an increase in
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!15
The number of observations for the regressions is substantially lower than the number of counts that link
to votes because observations with incomplete counts and questions for which member-level responses do
not exist are dropped from this part of the analysis. As the project continues, it may be possible to
incorporate additional observations where there are references to poll aggregates in the archival record, but
!
"*!
either the base or party size produces increases in the number of majority party votes cast
for the party position, although for both explanatory variables the magnitude of the
associated increase is substantially less than “one-to-one.” Interestingly, the only
statistically and substantively significant difference by period is for 1997-2002, the six
years of Republican control for which member-level whip data are available. Controlling
for the size of the base and – this is key – the size of the majority party, DeLay and his
colleagues were especially successful at growing the size of the majority party vote.
Further insight on the relative vote-growing performance of DeLay and other
whips is provided by the second regression equation, which drops the period dummies
and instead includes indicator variables for the individuals who served as majority whip.
None of the whip indicators achieve statistical significance (although Coelho/Gray is
statistically significant and DeLay comes close if we use a one-tailed test). Yet, the sizes
and signs of the parameter estimates for the whip indicators are instructive. The DeLay
period still stands out, but so does the session in which Coelho and then Gray served as
majority whip for the Democrats. Indeed, the substantively small size of the parameter
estimate for the 1983-94 period in equation 1 appears to result from the divergent vote-
growing experiences of Coelho and Gray, on the one hand, and Thomas Foley and David
Bonior, on the other.
The results for Coelho/Gray are not unexpected. Close observers of party politics
in the House have drawn parallels between Tom DeLay and Tony Coelho, just as they
have between Newt Gingrich, R-GA, and James Wright, D-TX, who served as speakers
when DeLay and Coelho, respectively, were majority whip. Like DeLay, Coelho earned
a reputation for aggressive coalition building rooted in his close links to outside interest
groups, lobbyists, and activists associated with his party (Barry 1990, Sinclair 2006).
And although Coelho stepped down as whip in June 1989 amid an ethical cloud, the basic
structure and personnel makeup of his whip operation remained in place for the
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
for now data about the size of the base is only used where it can be calculated with confidence from
individual level responses.
!
#+!
remainder of the 101st Congress, perhaps contributing to Gray’s initial success in the
post.16
4. Outcomes
The ultimate aim of majority whipping, of course, is not to gather votes but to win
on the floor. Indeed, once the 218-vote target has been achieved, there often are
incentives for the majority leadership to allow cross-pressured members to vote against
the leadership and with important interests in their districts, helping these members build
electoral support at home. We need to consider the outcomes of the questions that were
whipped by House majorities, 1955-2002. Overall, how successful was the majority
party on whipped matters? Are there significant differences across periods or individual
whips?
As part of the broader research for this project, detailed legislative histories were
constructed for the hundreds of measures and over 1,000 polled questions included in the
underlying data set. Considerable strategic information was gleaned from the archival
evidence. In constructing these legislative histories, I also leaned on the standard sources
used by congressional scholars. More concretely, the leadership papers were
supplemented with information from the Congressional Quarterly Almanac, CQ Weekly
Report, The Washington Post, and other national media outlets.17
When there were
ambiguities, I consulted the floor debate on the matter in the Congressional Record,
including any discussion of special rules and other procedural preliminaries. Based on
these materials, it was fairly straightforward to assign the outcomes for each polled
question to one of five main categories.
The first outcome category is for unambiguous wins by the majority party. Here,
no significant changes were made to broaden support for the underlying proposal or
motion and the majority party carried the day on the floor. During April 6-7, 1965, for
example, the Democratic leadership polled its members about their views on the GOP
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!16
Indeed, the master records for whip counts conducted during 1989 were located in the personal papers of
David Bonior, who served as chief deputy whip to both individuals, rather than the papers of either Coelho
or Gray. 17
Prior to the 1980s, CQ Almanac generally is sufficient, but after that point it was often necessary to also
consult the weekly reports for sufficiently detailed information about coalition building on whipped
questions.
!
#"!
motion to recommit and final passage for the legislation that established the Medicare
program. The motion to recommit failed and the bill passed. Both questions are treated
as unambiguous wins for the majority party. In late February 2001, the Republican
leadership asked GOP members if they would support the proposed Bush tax cut as
reported by the Ways and Means Committee. The bill was considered on the floor
without major change and passed. This item also is treated as a clear-cut majority party
win.
The second category is for questions that carried but only after major changes
were made. Obviously, few major bills are adopted on the floor without the passage of
any amendments at all. This is particularly the case for appropriations bills and other
omnibus measures, which can draw dozens of amendment proposals. So here, it was
crucial to distinguish between modifications based on their substantive importance and
impact on the distribution of viewpoints on the polled question. Since whip counts
typically deal with highly salient aspects of the legislative agenda and receive ample
media attention, it was not difficult to identify instances of major change between the
whip count on a matter and the decision made on the floor. On May 20, 1993, for
example, the Democratic leadership polled its members about the rule for and initial
passage of reconciliation language championed by the Clinton administration. Between
the beginning of the whip process and the relevant votes on May 27, however, the
administration was forced to accept major substantive changes relating to a proposed
energy tax and spending on entitlements. Otherwise the bill would not have passed. The
entitlement modifications were agreed to the day before the vote and folded into the
measure on the floor via the rule. As a result, both final passage and the rule are treated
as instances of major change.
Third is a category for items that the majority party opted to pull from the agenda.
Usually these decisions were made after the relevant whip count indicated significant
potential opposition within the majority caucus and there was little prospect of
countervailing support from within the minority party. In assigning whip questions to
this category, I looked for concrete evidence that the lack of floor action was a strategic
decision made by the leadership or the relevant committee chair, rather than an indirect
byproduct of some other action. On March 16, 1966 and again on September 1, 1967, for
!
##!
example, House Democratic leaders polled their colleagues about proposals to allow
common situs picketing, and in both cases the underlying measure was not brought
before the full House because the counts indicated insufficient support.
A fourth category is for unambiguous majority party losses. In June 1981, for
example, the Democratic leadership asked its members if they would support the
previous question on the rule for the reconciliation bill of that year. The rule would have
precluded Republicans from offering the Reagan budget cuts as a substitute. After a
sufficient number of Democrats joined Republicans to vote against the previous question
motion, the GOP offered an alternative rule that was adopted, which in turn enabled them
to secure floor action and eventual passage of the Reagan initiative. Here, both the
previous question motion and a related poll about passage of the Democratic bill are both
treated as unambiguous defeats for the majority party.
The final category, “other,” is for outcomes that could not be meaningfully coded
as wins, instances of major change, pulled items, or unambiguous losses. Usually, these
questions were obviated by events within the House or otherwise did not receive floor
attention for reasons rooted in the legislative context. On January 4, 1956, for example,
the House Democratic leadership asked rank-and-file members whether they would be
willing to support a highly salient school construction bill that had been reported the
previous year by the Committee on Education and Labor. The bill as reported did not
include controversial language drafted by Adam Clayton Powell, D-NY, that would have
barred school construction funds from school districts that practiced racial segregation.
When the measure was finally considered on the floor in July, the Powell amendment
was adopted and the bill as modified went down in defeat. Clearly, the outcome was far
from a victory for Democratic leaders, but at no point during floor action that year was
there a freestanding decision by the House on the measure as reported and that was the
gist of the polled question. As a result, this case is placed in the “other category” for the
purposes of analysis.
Table 7 summarizes the outcomes for over 1,000 questions whipped by the House
majority leadership across the four periods and using the aforementioned categorization
scheme. Included are the motions that could be directly linked to recorded votes and thus
constituted the evidentiary base for the analysis reported in Tables 5 and 6. But also
!
#$!
included are hundreds of questions that were not the subject of roll call votes and for
which outcome categories still can be assigned. There are some minor differences across
periods in the incidence of cases placed in the “other” category. This category comprised
less than 3 percent of observations for 1955-72, but rose to 8-9 percent for the each of the
next three periods. The cause may be the higher percentage of final passage questions for
the Textbook Era – Such questions are less likely to be obviated by events. But to
facilitate comparisons across periods, this category is excluded from Table 7.
Several patterns are discernable. For one, although the strategic calculations of
party leaders imply that poll questions should be “in play” and the outcome in serious
doubt – otherwise the leadership would not allocate valuable resources to the whipping
effort – across all four periods the whips won most of these contests. Even with the fairly
restrictive definition of party success employed here, the leadership unambiguously
prevailed on almost three-quarters of the polled questions. That said, if we combine the
categories for major change, pulled items, and unambiguous losses, the total is about 39
percent. Although the leadership success rates are high, in other words, they still tend to
lose (to varying degrees) on a substantial minority of whipped questions. Any broad
inferences about “party government” would be overdrawn.
Second, there are some noteworthy differences across periods. Most important,
the proportion of unambiguous losses declined substantially over time, from a high of
over 20 percent in 1955-1972 to just eight percent in 1995-2002. The real distinction,
however, is between the first two periods, taken together, and the latter two periods.
Indeed, the outcome distributions for 1955-72 and 1973-82 are very similar, as are the
distributions for 1983-94 and 1995-2002. Across the 1980s and 1990s, the most
prominent difference concerns the “major change” category, and here DeLay and his
colleagues were about twice as likely to be associated with major modifications in the
content of party initiatives as was the case for the Democratic whips of 1983-1994.
Once again, the main reason may have been the relatively small sizes of the
Republican majorities. As a result, I conducted a multivariate analysis of the outcomes of
the whip process across time periods and the results are reported in Table 8. Included are
the explanatory factors from the vote gathering analysis reported ion Table 6. Here, the
dependent variable is an ordinal ranking of the outcome category, with “win” as the
!
#%!
highest ranking, “major change” second, “pull” third, and “losses” at the low end of the
scale. Items placed in the “other” category are excluded. The estimator is ordered
probit.18
Once again, increases in the size of the base and of the majority party are
associated with positive outcomes for the whips. Whip success (defined as more
unambiguous wins and fewer instances of major change, pulled items, and clear-cut
losses) was significantly higher for 1983-94 and for 1997-2002 than for the earlier
periods. After controlling for party size, however, the relative effectiveness of the DeLay
whip operation is fully apparent.
5. Whip Counts and the Floor Median
Before concluding, in this section I look more closely at the poll behavior of
centrist legislators – members with preferences near the chamber median – because of the
pivotal role these members play in the leading theories of lawmaking. Characteristics of
their whip count behavior provide further guidance about how party theories should be
developed. The scholarly dispute about party influence in Congress is well known, but a
few details will help motivate the discussion in this section.
Krehbiel (1998) presents a spatial model of lawmaking without consequential
party organizations. As is standard with spatial theory, legislators are assumed to have
preferences that can be represented as ideal points along one or more underlying
dimensions of evaluation, typically the liberal-conservative continuum. Legislative
alternatives and the status quo (existing law or some other reversion point) are also
represented as points along the evaluation dimension. By assumption, legislators prefer
the alternative that is spatially most proximate to their ideal points and cast their votes
accordingly. If the focus is on a legislature with the internal rules and structure of the
U.S. House of Representatives, then the outcomes of chamber decision making should
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!18
The number of observations is substantially lower relative to Table 7 because observations with
incomplete counts and questions for which member-level responses do not exist are dropped from the
analysis. Moreover, it was not feasible to conduct a second regression with indicator variables for
individual whips. The results of such an analysis were not robust to alternative specifications, probably due
to the number of categories in the dependent variable and the large number of dummy variables on the
right-hand side.
!
#&!
approximate the ideal point of the floor median, the only alternative that cannot be
successfully amended. Party organizations and leaders are not outcome consequential.19
The cartel model of Cox and McCubbins (and here I am mostly referring to the
version in their 2005 book) is rooted in the same spatial logic as Krehbiel’s theory, but
includes a majority leadership with agenda-setting powers. If the leadership places a bill
on the agenda, then the outcome of the House legislative process will tend to approximate
the preferences of the chamber median a la the “weak parties model. But if a majority of
the majority caucus prefers the status quo of existing law to the position of the floor
median, then the leadership will close the gates and keep the bill off the agenda.
Although the bills that pass reflect centrist preferences within the chamber, the majority
leadership influences policy outcomes by blocking changes that would make a majority
of the majority party worse off.
The conditional party government argument advanced by Aldrich and Rohde
(Rohde 1991, Aldrich and Rohde 1998) also shares much of the basic spatial logic of the
other theories, including member ideal points that are exogenously determined, a single
underlying dimension of evaluation (typically, although not always), and preferences
over alternatives based on spatial proximity. However, according to this model, when
preferences within the majority party are homogeneous and the distance between the two
party medians is large, rank-and-file members of the majority caucus will take steps to
strengthen their leaders, enabling them to regularly pass bills that diverge from the
position of the floor median toward the majority party program.
None of these theories are consistent with key aspects of the majority whip
process in the House. Perhaps a simple thought experiment would help clarify matters.
Suppose that the majority leadership has allowed a bill to be placed on the House agenda.
The measure has been marked up in committee, perhaps amended in the Committee of
the Whole, and now awaits a roll call on passage. Following the floor amendment
process and immediately prior to the final passage vote, the majority leadership conducts
a whip count, asking majority party members whether they will vote “yes” on initial
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!19
Krehbiel’s theory incorporate supermajority procedures like the Senate filibuster and the 2/3 vote
necessary to override a presidential veto, which of course can produce noncentrist outcomes. However, if
we focus on initial House passage of legislation and set aside certain strategic calculations (e.g., adjusting
legislation early on to secure maximum support and perhaps forestall a veto), then the prediction of median
outcomes is appropriate for models of the House without parties.
!
#'!
House passage. No significant modifications are made to the legislation between the time
of the whip count and the roll call on final passage. The bill clears the House. Under this
scenario, would we expect large number of centrist legislators with preferences near the
position of the floor median to respond as “undecided” on the whip poll? The answer to
the question varies, depending on which theory of lawmaking is under consideration.
If we embrace the central premises of the spatial model, and we assume that
leadership influence is absent from the legislative process or is exercise primarily at the
agenda setting stage, then the expectation is that bills passing the House should fully
reflect the preferences of the floor median. The assertion follows directly from the “weak
parties” and party cartel models. As a result, if the aforementioned whip count concerns
initial passage of a measure that is indeed adopted by the chamber without intervening
changes, then members with preferences located near the floor median should be
disproportionately less likely to respond as undecided or in opposition on the poll.
Instead, the underlying spatial logic implies that undecided lawmakers should be
concentrated near the cutting point that is equidistant between the final version of the bill
and the status quo of existing law, wherever that might be located. But the floor median
– who according to the majoritarian and cartel models is getting exactly the legislation
she most prefers – should be if anything less likely than the typical legislator to respond
as “undecided” or opposed on the initial passage whip count.
For the right set of polled questions and roll calls, then, the whip count responses
of centrist legislators may illuminate the explanatory power of the competing theories of
lawmaking. Again, it is critical that we focus solely on whip counts about initial House
passage on measures that actually cleared the chamber without significant intervening
modifications. None of the theories predict that amendments, procedural motions, or
failed bills will necessarily reflect the floor median position – only bills that pass. And
the leadership needs to have conducted a poll that explicitly asked for member views
about the full content of the measure that cleared the chamber. Questions about
particular titles, contingencies, or parts of the legislation are not sufficient here.
Fortunately, the scenario of my thought experiment is not all that contrived. More than a
hundred questions polled by the House majority party, 1955-2002, fit the bill.
!
#(!
To make the analysis tractable, I focus on the subset for 1955-62, 1978-84, and
1997-2002, roughly representing the beginning, middle, and end of the time span covered
in this research project. Included are ten items from 1955-62, 31 for 1979-84, and 55 for
1997-2002. To identify which members have ideological views near the chamber
median, I rely on first-dimension DW-NOMINATE values.20
Especially for the 1950s-
1970s, many votes on Capitol Hill were multidimensional (Poole and Rosenthal, 1997).
But the leading spatial theories generally posit a unidimensional choice space and first-
dimension NOMINATE has structured party differences in the roll call record for more
than a century. For the exploratory purposes of this section, first-dimension NOMINATE
data can help us approximate the subset of members near the effective floor median in a
systematic fashion. To identify them, all House members during the appropriate
Congresses were rank-ordered by their NOMINATE values and then (based on the
ordering) divided into deciles of approximately 43-44 legislators each, with members of
the fifth and sixth ideological groups treated as near the collective floor median for the
purposes of analysis.
The results are summarized in Table 9. The ideological decile in which a member
falls (for the full chamber) is denoted in the first column, and then for each subset of the
evidence, denoted are the percentage of the poll respondents answering as “yes or leaning
yes,” “undecided,” and “no or leaning no.” If the simple majoritarian version of the
“weak parties” perspective and the party cartel theory are helpful for understanding the
whip process, then the percentage of undecided and of opposing members should not
peak at the fifth and sixth ideological deciles. Again, why would a House member who
had just carried the day on one of the more salient measures on the floor agenda respond
as undecided or opposed regarding her own most-preferred policy outcome?
Interestingly, the percentage of undecided and opposed legislators peaks
substantially to the ideological right of the floor median for 1978-84. These were
transitional congresses between the large, highly divided partisan majorities of the 1970s
House and the more polarized chamber of recent years and centrist legislators may have
indeed driven the lawmaking process. However, for the small number of whipped items
in 1955-62 and the very large number from the DeLay years, the likelihood of being
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!20
These data were downloaded from http://voteview.com.
!
#)!
undecided or opposed was indeed highest near the ideological views of the chamber
median.
Keep in mind that for the late 1950s and early 1960s, the scope of whip activity
was narrow and highly selective, and as indicated throughout this paper, the majority
leadership of that era was rolled on the floor relatively frequently. Still, for the small set
of items from those years upon which the Democratic whip network was fully engaged
and – this is key – ultimately successful, the results of Table 9 are inconsistent with the
majoritarian version of the spatial model and with theories positing parties that primarily
exert influence through agenda setting. And most important, for the very large number of
analogous Republican items from the late 1990s and early 2000s, far more pervasive
inconsistencies are apparent between these conceptual constructs and the results of Table
9.
Of course, an alternative explanation is that chamber centrists were often
undecided or opposed because the bills diverged from their policy preferences toward the
majority party position a la conditional party government. Perhaps the status quos were
located somewhere on the minority party’s side of the floor median, the bills diverged
from that median in the direction of most majority party members, and the views of
centrist legislators were located somewhere in the middle, near the relevant cutting line.
Cross-pressured between two alternatives roughly the same distance from their ideal
points, not surprisingly the centrists often were undecided or opposed about final passage
just days or hours before the vote. On first blush, this scenario seems plausible, at least
for the small subset of items considered for 1955-62 and the large set for 1997-2002.
However, there also are good reasons not to fully embrace the standard
conditional party government explanation. For one, any party influence detected during
the 1950s and 1960s would have occurred before the party building reforms of the 1970s
and 1990s, which are central features of the conditional party government argument.21
Moreover, the spatial representation of conditional party government posits member
preferences that are fixed and determined exogenously to the legislative process,
including leadership lobbying on the floor. How can this perspective inform the process
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!21
But see Lee (2009) for a “team-based” conceptualization of party influence that is less reliant on formal
prerogatives.
!
#*!
through which members form preferences when preferences are treated as exogenous?
More generally should party influence even be conceptualized and measured in terms of
some deviation from the preferences of the floor median toward median preferences
within the majority party? And what precisely is the median preference on the floor
when dozens of centrist members are “undecided” on major roll calls just days before
votes occur?
6. Conclusion
My analysis of nearly fifty years of majority whip history in the House indicates
that, under the right conditions, the process of endgame lobbying by the leadership has
significant consequences for the composition of coalitions and the content of legislation.
The quantity of whip activity has increased substantially since the mid-1950s, largely
because of heightened partisan polarization within the chamber and expansions in the
floor workload. Although whip activity has predominantly targeted final passage votes
and entire bills for many decades, the kinds of questions that were polled did change over
time in response to the partisan and procedural context. When presidential vetoes were a
central feature of party politics in the 1970s, for example, the majority leadership often
polled on veto overrides, and as the chamber turned toward increasingly restrictive
amendment rules in the 1970s and 1980s, the percentage of whip activity allocated to
procedural matters tripled for a time. For the most part, the substantive issues targeted
for whip activity track the broader policy agenda in important ways, but especially in
recent years appropriations bills and other managerial challenges have been prominent on
the whipped agenda.
Although the size of the leadership’s base (“yes” plus “leaning yes” responses on
whip counts) has increased over time as the House became more polarized along partisan
lines, on whipped items the majority leadership still typically needs to pick up the support
of 30 or 40 of their fellow partisans in the days prior to major floor votes, and also
usually relies on nontrivial support from minority party members in order to prevail.
There is a surprisingly amount of position movement on even the most salient features of
the legislative agenda and toward the end of the process of floor decision making. Across
the four time periods that structure much of this study, a systematic analysis of the
!
$+!
outcomes of polled questions indicates that the partisan majority usually wins. Still, the
proportion of unambiguous victories for the whips increased over time with party
polarization, from 65 percent or so during 1955-72 and 1973-84 to almost 80 percent
during the 1980s and 1990s. Similarly, the percentage of unambiguous leadership losses
during the 1980s and 1990s was about half the level that characterized earlier decades.
After controlling for the size of the base and of the majority party, the whip
operations of Tom DeLay during the GOP era appear to have been remarkable effective,
although there are indications of similar effectiveness for Democratic whips Tony Coelho
and William Gray, at least for the available evidence from the first session of the 101st
Congress. Overall, the whip successes of the 1983-94 period may have mostly derived
from increased preference homogeneity within the party (as reflected in part in the size of
the base on whip counts) and the size of the majority party caucus, whereas the
independent impact of the whips and other party leaders seems especially consequential
for the Republican majorities of 1995-2002.
Much of the analysis of majority whip operations in this paper resonates more
strongly with the conditional party argument (CPG) than it does with the party cartel
theory or conceptualizations of lawmaking that do not integrate partisan institutions and
incentives in important ways. When the majority party is relatively unified on the major
issues of the day and there is a substantial gulf between the preferences of majority and
minority party members, CPG maintains, then the majority rank and file will tend to
empower their leaders with the resources necessary to whip potential defectors into line.
Indeed, as the chamber became more divided along party lines in the 1980s and 1990s,
the impact of the whip operations seemed to rise. Especially given the narrow partisan
majorities of the Republican era – indeed, perhaps because of them – Tom DeLay’s vote
gathering efforts were particularly impressive and appear to mesh well with the
conditional party government argument.
Still, my analysis of five decades of whip history also indicates that much of what
constitutes party influence in the House concerns the formation of positions and
preferences and that appropriate conceptualizations and empirical tests of party influence
need to address the emergence and movement of member positions. As a result, a more
dynamic version of the conditional party government argument, rooted in the process
!
$"!
through which members made up their minds about major issues, may better account for
the rhythms of whip activity in the House. For one, the distribution of member
preferences or positions should not be treated as exogenous to the process of party
influence. Instead, the leadership will gauge the evolving positions of rank-and-file
members prior to floor action, focusing its attention primarily on items important to the
party’s name brand and for which there is at least the possibility of intra-party unity.
Depending on the preference context, the leadership uses a menu of tactics – retreat,
compromise over legislative substance, the insertion or removal of issue dimensions, side
payments, persuasion, and so on – as part of efforts to build a winning coalition. The
likelihood that the majority party will prevail should depend on the size of its base, the
extent to which the opposition is entrenched or positions are in flux, the feasibility of
altering the dimensional structure of deliberation or significant increasing support via
incremental modifications, and the availability of side payments. Obviously, the size of
the majority caucus is a central factor. But the critical departure is to construct theories
of party behavior that capture the impact of the whips and other leaders on the process of
position and preference formation over time.22
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!22
See Smith (2007), Krehbiel (1991), Arnold (1990), Lee (2009), and especially Cooper and Hering (2003)
about the importance of integrating preference endogeneity and position formation into theories of
lawmaking.!
!
"#!
Figure 1. Majority party whip activity, 1955-2002
!
""!
Figure 2 Majority party cohesion, votes and whip polls
!
"#!
Table 1. Types of “questions” polled, by period
1955-72 1973-82 1983-94 1995-2002 Total
Bills, general
policy
121
(72.46)
131
(49.25)
132
(44.00)
242
(55.13)
626
(53.41)
Conference
matters
6
(3.59)
18
(6.77)
22
(7.33)
35
(7.97)
81
(6.91)
Amendments,
parts of bills
24
(14.37)
59
(22.18)
88
(29.33)
93
(21.18)
264
(22.53)
Motion to
recommit
5
(2.99)
6
(2.26)
5
(1.67)
10
(2.28)
26
(2.22)
Procedure,
strategy
6
(3.59)
36
(13.53)
51
(17.00)
59
(13.44)
152
(12.97)
Veto
overrides
5
(2.99)
16
(6.02)
2
(.67)
0 23
(1.96)
Total
167 266 300 439 1172
!
"#!
Table 2. Prevalence of question types for nonunanimous votes by whip status
1955-72
1973-82
1983-94
1995-2002
Total
Whip 60.83 44.13 37.89 53.27 48.29 Bills, general
policy Other 35.52 25.21 17.75 12.43 21.68
Whip 4.17 8.94 9.69 11.21 9.33 Conference
matters Other 14.04 8.69 8.18 5.80 8.70
Whip 11.67 20.67 29.96 21.50 22.20 Amendments,
parts of bills Other 19.43 49.37 43.13 46.77 41.97
Whip 13.33 3.91 2.20 1.87 4.01 Motion to
recommit Other 16.13 3.38 3.24 5.20 6.04
Whip 3.33 13.41 17.62 10.90 12.16 Special Rules
Other 7.07 6.03 11.64 13.16 9.53
Whip 2.50 5.59 .88 0 1.77 Veto
overrides Other .47 .32 .41 .40 .39
Whip 4.17 3.35 1.76 1.25 2.24 Other
Other 7.33 7.00 15.65 16.24 11.70
Note: cell entries are the percentage of observations for the time period falling in the
relevant category (all nonunanimous votes, whipped versus non-whipped items). Roll
calls for 1976, 1987-88, and 1991 are excluded because of missing whip data.
!
"#!
Table 3. Prevalence of policy areas for nonunanimous votes by whip status
1955-72 1973-82 1983-94 1995-2002
Whip 11.67 14.53 13.66 11.21 Congress,
government
operations Other 16.87 11.72 15.87 20.82
Whip .83 2.23 10.57 3.12 Defense,
homeland
security Other 4.82 5.45 8.80 5.70
Whip 9.11 8.94 9.25 4.05
Foreign policy Other 7.39 8.54 8.21 5.37
Whip 20.00 24.02 20.26 17.76 Economy,
budget,
taxation Other 8.64 14.54 9.13 11.57
Whip 5.83 1.68 2.20 2.18
International
trade Other 2.78 1.84 2.25 1.72
Whip 5.0 14.53 1.32 4.05
Energy,
environment Other 6.91 11.83 7.81 5.54
Whip 6.67 1.68 1.32 0
Civil rights Other 6.02 1.84 1.10 .60
Whip 7.5 7.82 3.52 3.12
Social welfare Other 5.87 4.46 4.16 2.92
Whip 1.67 1.68 1.32 4.36
Health Other .58 1.05 2.03 2.12
Whip 7.50 0 3.08 3.43
Education Other 3.14 1.6 2.47 2.42
Whip 5.83 7.82 7.49 4.67
Labor,
consumers Other 4.03 3.64 1.80 1.43
Whip 5.00 5.03 18.94 35.51
Appropriations Other 18.39 20.69 27.07 31.99
!
"#!
Table 4. Distribution of poll responses on completed counts
1955-72
1973-82
1983-94
1997-2002*
Yes
56.38
50.40
58.81
57.57
Leaning yes
2.09
8.16
10.30
16.32
Undecided
13.39
17.51
15.91
11.25
Leaning no
1.05
3.04
2.76
3.85
No
11.57
8.94
5.37
4.13
Other
15.51
11.95
6.84
6.88
* In addition to portions of 1976 and 1987-88, member level records are unavailable for
most of the majority party counts conducted during 1995-96.
!
"#!
Table 5. Majority party base, vote pickup, and minority support
1955-72
1973-82
1983-94
1995-2002
Mean base (yes plus leaning
yes), nonunanimous votes
148.22 157.52 179.20 165.25
Mean base (yes plus leaning
yes), party-line votes
150.32 163.81 182.99 172.23
Mean majority pickup,
nonunanomous votes
36.98 34.70 27.74 30.27
Mean majority pickup, party-
line votes
36.46 33.90 27.83 30.76
Mean minority support,
nonunanimous votes
41.84 40.26 35.18 42.29
Mean minority support, party-
line votes
30.63 28.62 24.92 21.64
!
"#!
Table 6. Majority vote gathering success 1955-2002
(1) (2)
majority majority
votes votes
base 0.638*** 0.656***
party size 0.243*** 0.280**
1971-82 -0.487
1983-94 1.263
1997-2002 7.921*
Boggs (1962-70) -2.951
O’Neill (1971-72) 2.975
McFall (1973-76) 1.768
Brademas (1977-80) -3.487
Foley (1981-86) -2.047
Coelho/Gray (1987-91) 9.678
Bonior (1991-94) -5.773
DeLay (1995-2002) 7.435
Constant 29.658 18.744
Observations 561 561
R-squared 0.433 0.443
* significant at 10%; ** significant at 5%; *** significant at 1%
!
"#!
Table 7. Majority Party Success on polled items, 1955-2002
1955-72
1973-82 1983-94 1995-2002 Total
Win
93
(65.03)
151
(65.65)
214
(79.85)
299
(77.46)
757
(73.71)
Major
change
11
(7.69)
21
(9.13)
11
(4.10)
29
(7.51)
72
(7.01)
Pull
10
(6.99)
13
(5.65)
15
(5.60)
27
(6.99)
65
(6.33)
Loss
29
(20.28)
45
(19.57)
28
(10.45)
31
(8.03)
133
(12.95)
Total
143
230
268
386
1027
!
"#!
Table 8. Outcomes for the majority party on polled items (ordered probit, cell
entries are marginal effects)
(1) (2) (3) (4)
win change pull loss
base .00266*** -.00045*** -.00054*** -.00166***
party size .00179 -.0003 -.00036 -.0011
1973-82 -.0304 .0050 .0062 .0192
1983-94 .0956* -.0172* -.0201* -.0582*
1997-2002 .146** -.0274** -.0313** -.0872**
Observations; 661
Cutpoints statistically significant
McKelvey & Zaviona’s R2; .110
significant at 10%; * significant at 5%; ** significant at 1%***
!
"#!
Table 9. Whip count support and opposition by ideological deciles on successful
final passage questions, selected congresses
1955-62 (10 questions)
1979-84 (31 questions) 1997-02 (55 questions)
decile
Y/LY U N/LN Y/LY U N/LN Y/LY U N/LN
1
90.89 7.09 2.03 79.48 14.44 6.08
2
89.03 7.98 2.99 85.04 11.12 3.84
3
74.81 18.83 6.36 81.29 13.64 5.07
4
70.68 20.42 8.90 68.29 22.21 9.50
5
53.71 26.50 19.79 57.14 30.66 12.20 64.63 23.41 11.96
6
26.11 35.47 38.42 43.96 36.66 19.38 75.65 17.53 6.82
7
21.52 25.32 53.16 22.32 41.67 36.01 85.95 10.94 3.11
8
8.11 29.73 62.15 17.14 40.00 42.86 87.82 8.85 3.33
9
0 18.75 81.25 39.13 30.43 30.43 86.80 8.73 4.47
10
5.41 45.95 48.65 78.90 11.34 9.76
NOTE: Cell entries are the percentage of all respondents providing a substantive
response (yes to no) that fell within each position category.
!
"#!
References
Aldrich, John H., and David W. Rohde. 1998. “Measuring Conditional Party
Government.” Presented at the annual meeting of the Midwest Political Science
Association, Chicago.
Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven: Yale
University Press.
Barry, John. 1990. The Ambition and the Power. New York: Penguin.
Black, Duncan. 1958. The Theory of Committees and Elections. Cambridge:
Cambridge University Press.
Burden, Barry D., and Tammy M. Frisby. 2004. “Preferences, Partisanship, and Whip
Activity in the House of Representatives.” Legislative Studies Quarterly: 29, 569-90.
Clark, Champ. 1920. My Quarter Century in American Politics. New York: Harper &
Brothers.
Cooper, Joseph, and David W. Brady. 1981. “Institutional Context and Leadership
Style: The House from Cannon to Rayburn.” American Political Science Review 75:
411-25.
Cooper, Joseph, and Martin Hering. 2003. “Proximity Voting Versus Party Effects in
Congressional Decision Making.” Unpublished manuscript, The Johns Hopkins
University.
Cox, Gary W., and Mathew D. McCubbins. 1993. Legislative Leviathan: Party
Government in the House. Berkeley: University of California Press.
Cox, Gary W., and Mathew D. McCubbins. 2005. Setting the Agenda: Responsible
Party Government in the U.S. House of Representatives. Cambridge: Cambridge
University Press.
DeLay, Tom. 2007. No Retreat, No Surrender: One American’s Fight. New York, NY:
Penguin Group.
Dodd, Lawrence C. 1978. “The Expanded Roles of the House Democratic Whip
System: The 93rd
and 94th
Congresses.” Congressional Studies 7: 27-56.
Dodd, Lawrence C., and Terry Sullivan. 1981. “Majority Party Leadership and Partisan
Vote Gathering: The House Democratic Whip System.” In Understanding
Congressional Leadership, ed. Frank H. Mackaman. Washington, D.C.: CQ Press.
Evans, C. Lawrence and Claire E. Grandy. 2009. “The Whip Systems of Congress.” In
!
""!
Congress Reconsidered, 9th ed., ed. Lawrence C. Dodd and Bruce I. Oppenheimer.
Washington, D.C.: CQ Press, 189-216.
Evans, C. Lawrence. 2011. The Whip Systems of Congress. Unpublished manuscript,
The College of William and Mary.
Jackson, John E., and John W. Kingdon. 1992. “Ideology, Interest Group Ratings, and
Roll Call Votes.” American Journal of Political Science 36: 805-23.
Kingdon, John W. 1973. Congressmen’s Voting Decisions. New York: Harper and
Row.
Krehbiel, Keith. 1991. Information and Legislative Organization. Ann Arbor, MI: The
University of Michigan Press.
Krehbiel, Keith. 1998. Pivotal Politics: A Theory of U.S. Lawmaking. Chicago, IL:
University of Chicago Press.
Lee, Frances E. 2009. Beyond Ideology: Politics, Principles, and Partisanship in the
U.S. Senate. Chicago, IL: University of Chicago Press.
Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven, CT: Yale
University Press.
Mayhew, David R. 2011. Partisan Balance. Princeton, N.J.: Princeton University
Press.
Polsby, Nelson W., and Eric Schickler. 2002. “Landmarks in the Study of Congress
since 1945.” Annual Review of Political Science 5: 333-67.
Poole, Keith T. and Howard Rosenthal. 1997. Congress: A Political-Economic History
of Roll Call Voting. Oxford: Oxford University Press.
Ripley, Randall B. 1964. “The Party Whip Organization in the United States House of
Representatives.” American Political Science Review 58: 561-76.
Rohde, David W. 1991. Parties and Leaders in the Postreform House. Chicago, IL:
University of Chicago Press.
Sinclair, Barbara. 1983. Majority Leadership in the U.S. House. Baltimore, MD: Johns
Hopkins University Press.
Sinclair, Barbara. 1995. Legislators, Leaders, and Lawmaking: The U.S. House of
Representatives in the Postreform Era. Baltimore, MD: Johns Hopkins University Press.
!
"#!
Sinclair, Barbara. 2006. Party Wars: Polarization and the Politics of National Policy
Making. Norman, OK: University of Oklahoma Press.
Smith, Steve S. 2007. Party Influence in Congress. Cambridge: Cambridge University
Press.
Theriault, Sean. 2008. Party Polarization in Congress. Cambridge: Cambridge
University Press.